A statistical study on how the new system changes member engagement
Wellspring is a Canadian cancer support organization offering free programs to anyone affected by cancer. In March 2024, they introduced a new registration system to improve accessibility and engagement.
This project explores how factors like parenthood, gender, and personal characteristics relate to program participation, using real member and attendance data.
Our goal is to help Wellspring understand what drives program engagement and how to better support diverse members.
The population of interest includes all individuals who registered with Wellspring and who participated in, or were scheduled to participate in, at least one program between 2023 and early 2024.
This includes members of all ages, genders, and backgrounds across Canada, regardless of cancer type, who accessed Wellspring services either online or in person.
Our analyses focus on observed behaviors and characteristics from this population, as recorded in Wellspring’s program registration and attendance system.
This project uses two datasets:
Most members are from Ontario, with the most common cancer type reported being breast cancer.
Engagement varies widely: the median number of sessions attended by members is 2, but some attended more than 250.
Many fields, including LGBTQ+ status, impairments, and program interests, have high rates of missing data.
Variables:
How do age, number of program interests, and membership duration jointly predict total services attended?
This question is addressed by linear regression, as it allows us to model the simultaneous influence of multiple continuous predictors on a continuous outcome.
lm() function).Question: Do members with children under 18 attend Wellspring services more or less often than those without children?
Relevance: Understanding attendance patterns based on parental status helps Wellspring identify potential barriers for parents and adjust services accordingly. Examples include offering childcare at events or making events more child-friendly.
| Parent Status | Avg. Attended | Total |
|---|---|---|
| No | 70.63 | 6039 |
| Unknown | 38.43 | 31045 |
| Yes | 46.05 | 9219 |
Null hypothesis (\(H_0\)): There is no difference between the two groups
Alternative hypothesis (\(H_A\)): There is a difference between the two groups
Using a p-value less than 0.001 from two-proportion testing, there is strong evidence of a difference in attendance between members with children under 18 and those without.
We assume these data are representative.
The box-and-whisker plot shows that attendance for members with children tends to be lower than for members without children.
The median number of services attended by members with children under 18 is lower than that of members without children, and their attendance values are more concentrated around the median.
Limitations: some members chose ‘Unknown’ for parental status; this may affect the variability shown in Table 1.
Other factors, such as work schedules, motivation, and health, may also influence the results but are harder to quantify.
Conclusion: Members with children under 18 are less likely to participate in these services compared to members without children.
How has the new registration system affected engagement by gender?
We chose this research question because:
It is a two-parameter hypothesis testing question.
It provides Wellspring with insight into how the registration system is functioning.
The data we get from performing this two parameter hypothesis test will help Wellspring see if the new registration format is affecting client engagement by gender: if there exists a difference between engagement of female clients and male clients. Other gender categories were excluded due to insufficient sample sizes for reliable analysis. Graphs: Table of focused variables from data
| Gender | # | Total present | Total services | Proportion of attended |
|---|---|---|---|---|
| Female | 2519 | 25621 | 40115 | 0.6386888 |
| Male | 514 | 3503 | 4955 | 0.7069627 |
| NA | 1794 | 5985 | 10067 | 0.5945167 |
From the simulations we ran:
Calculated test statistic for the original data is 0.0682739.
We can see from the boxplot that the distribution is centered around 0.07–0.08. Thus the p-value is large (not < 0.05), and we cannot reject the null hypothesis of no difference in engagement between male and female members for the new registration format.
We must consider possible type II error and the impact of missing or unspecified gender data, as these factors could affect the results.
Conclusion: Based on the test statistics, there is no statistically significant difference in engagement between male and female members after the new registration format was introduced.
Re-summary
RQ1: Members who are younger, have more program interests, and have been registered longer tend to attend more services overall.
RQ2: Members with children under 18 attended fewer services on average than members without children, suggesting family responsibilities may limit participation.
RQ3: After the new registration system was introduced, engagement among different genders shifted slightly, with a modest improvement among female-identifying members.
Personal and social factors such as family obligations and gender interact with system-level changes (like registration updates) to shape engagement.
Structural improvements like system redesign can help. But their impact may vary across subgroups.
Program interest is a strong predictor of engagement, suggesting that initial onboarding and motivation may be key.
Collect additional variables on employment, caregiving, and scheduling availability to better explain participation gaps.
Conduct subgroup analyses across other demographics (e.g., age × gender).
Track long-term engagement trends to assess retention beyond the early months.